shrivastava95 / poc_ocr

A repository for the POC for parsing translation dicitonaries using OCR. This is a part of the C4GT work under ai-tools/dictionary-augmented-transformers.

poc_ocr

A repository for the POC for parsing translation dicitonaries using OCR. This is a part of the C4GT work under ai-tools/dictionary-augmented-transformers.

Requirements

• Python version 3.11
• ironpdf - pip install ironpdf
• ensure Pillow library is installed with version < 10 (deprecation errors)

Files and Folders

• page_boxes - stores the bounding box data for each page.
• page_images - stores the images of each page for pre-processing.
• page_outputs - stores the OCR outputs on each page.
• Hanuman_Chalisa_In_Odia.pdf - a sample pdf to test the OCR on.
• poc_ocr_window.py - the final POC for loading PDFs and showing the bounding boxes.
• poc_ocr.py - a fallback checkpoint for poc_ocr_window.py - to where only the pdf is loaded and there is no display.
• test.py - testing some features

Instructions to run

python poc_ocr_windows.py to run the POC. Once the window opens, load a PDF and then select a page to be displayed. Then view the page with the bounding boxes overlaid. If you spot any errors with the bounding boxes displayed for a particular page, select another PSM mode using the drop down menu and re-run the parsing for that page.

demo video - Instructions

This demo video should clarify instructions on how to use the POC. demo video - youtube